Performance Study in Heterogeneous Vehicular Networks Using Dual Connectivity and Deep Reinforcement Learning

Vehicular networks are a vast and significant research subject that attracts the transportation and telecommunications industries. The transportation industry’s primary goal is to ensure road safety. For this purpose, as a solution, it was necessary to combine cellular networks and vehicular technologies, which led to Cellular Vehicle-to-Everything (CV2X). Thus, vehicles that support 4th generation (4G) or 5th generation (5G) capacities must be capable of transparently adjusting to the existing dual infrastructure by enforcing the new 5G radio and current 4G Long-Term Evolution (LTE) technology. In this context, Dual connectivity (DC) was invented to enable vehicles to be connected simultaneously using the two technologies and to resolve the problem of frequent Handovers (HO). This study aims to enhance the network’s performance by reducing HO’s number. To do this, we opted to use and test our network with two Deep Reinforcement Learning (DRL) algorithms: Proximal Policy Optimization (PPO) and Advantage Actor-Critic (A2C). Simulation results and the comparison between these two algorithms reveal that PPO is the algorithm that performs the best in almost all scenarios.